Pure Rust + CUDA LLM inference engine

A simple, fast and robust program-aware agentic inference system.

ForeAct: Steering Your VLA with Efficient Visual Foresight Planning

Triton-based implementation of Sparse Mixture of Experts.

Agent World Model: Infinity Synthetic Environments for Agentic Reinforcement Learning

Helpful kernel tutorials and examples for tile-based GPU programming

Reinforcement Learning via Self-Distillation (SDPO)

OpenResearcher: A Fully Open Pipeline for Long-Horizon Deep Research Trajectory Synthesis

Modalities, a PyTorch-native framework for distributed and reproducible foundation model training.

The best ChatGPT that $100 can buy.

FP16xINT4 LLM inference kernel that can achieve near-ideal ~4x speedups up to medium batchsizes of 16-32 tokens.

My Solution and Notes for the Stanford CS336: LLM from scratch

This is the github repo for our ICLR 2026 paper "Improving Attributed Long-form Question Answering with Intent Awareness".

个人论文阅读笔记归档｜Paper Reading Notes

An AI-Native Platform for Benchmarking SRE Agents

PET: Optimizing Tensor Programs with Partially Equivalent Transformations and Automated Corrections

Lightweight coding agent that runs in your terminal

Wan: Open and Advanced Large-Scale Video Generative Models

Learning TileLang with 10 puzzles!

Fast, Sharp & Reliable Agentic Intelligence

Moonshot's most powerful model

Outcome driven agent development framework that evolves

On demand communication

Tiga: Accelerating Geo-Distributed Transactions with Synchronized Clocks

Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞

Visual Causal Flow

Official Repository of VisGym: Diverse, Customizable, Scalable Environments for Multimodal Agents